Ryan Michael Kohlbrenner MD, Presenter: Nothing to Disclose

Aliya Qayyum MBBS, Abstract Co-Author: Spouse, Employee, Imorgon Medical

David E. Avrin MD, PhD, Abstract Co-Author: Stockholder, Reed Elsevier

The small bowel is time-consuming to track and evaluate on CT and is challenging to automatically segment. The purpose was to determine the feasibility of small bowel tracking by developing a computerized algorithm that segments the small bowel lumen and generates its centerline as a reference for creation of reformatted images in cross section to the lumen.

A computerized scheme was developed using the Insight Segmentation and Registration Toolkit. Three CT scans with well-distended small bowel were included in the study, including one CT enterography and two CT scans with small bowel obstruction. Beginning at a user-defined start point within the bowel lumen, a region-growing segmentation algorithm was used to accumulate neighboring intraluminal voxels of air or fluid attenuation until a user-defined endpoint within the lumen was reached. The distance between the two points was arbitrary, assuming loops of bowel between the points remained least partially distended. A secondary segmentation of the extraluminal abdominal contents was used to further refine the binary results by subtraction. The Euclidean distance transform of the binary segmentation volume was then generated (characterized by high values near the luminal center, low values at the periphery, and zero values for extraluminal voxels). Employing the transform as a cost function, where high values correspond to low cost, the minimal cost path between the two seed points was determined. The approximate centerline was superimposed on the original CT volume, and a series of images were generated perpendicular to the centerline.

Transluminal images perpendicular to the centerline were successfully generated for all cases in the test series, using maximally spaced user-defined seed points within the distended small bowel lumina. The line connecting the seed points remained within the small bowel lumen for each case.

Small bowel luminal segmentation can be performed in cases of partially or fully distended small bowel. Automated centerlining of the segmentation result facilitates generation of transluminal images, which would improve evaluation of small bowel pathology.

Generating an enteric centerline and perpendicular images would improve localization and evaluation of small bowel pathology such as enteritis in Crohn disease, small bowel tumors, and the transition point in small bowel obstruction.

Kohlbrenner, R, Qayyum, A, Avrin, D, Semi-automated Small Bowel Segmentation with Automated Centerlining on Thin-slice CT.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045612.html